Hugging Face's logo

software-mansion
/
react-native-executorch-gemma-4-multimodal

Audio-Text-to-Text
ExecuTorch
google
deepmind
gemma4
edge
Model card Files
xet
 Community
NorbertKlockiewicz commited on
Commit
da20400
·
1 Parent(s): e488770

Backfill constant metadata method values in config.json from .pte

Browse files
Files changed (3) hide show
  1. e2b/mlx/config.json +58 -0
  2. e2b/vulkan/config.json +66 -0
  3. e2b/xnnpack/config.json +64 -0
e2b/mlx/config.json ADDED
@@ -0,0 +1,58 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
3
+ "model": "gemma_4",
4
+ "family": "gemma",
5
+ "capabilities": [
6
+ "text-generation",
7
+ "image-text-to-text",
8
+ "audio-text-to-text"
9
+ ],
10
+ "backend": "mlx",
11
+ "license": "gemma",
12
+ "variants": [
13
+ {
14
+ "file": "gemma4_e2b_mlx_int4.pte",
15
+ "precision": "int4",
16
+ "quantized": true,
17
+ "default": true,
18
+ "methods": {
19
+ "enable_dynamic_shape": true,
20
+ "get_bos_id": 2,
21
+ "get_eos_ids": [
22
+ 1,
23
+ 106,
24
+ 50
25
+ ],
26
+ "get_max_context_len": 2048,
27
+ "get_max_seq_len": 511,
28
+ "get_vocab_size": 262144,
29
+ "has_ple": true,
30
+ "image_placeholder_id": 262145,
31
+ "use_kv_cache": true,
32
+ "audio_encoder": {
33
+ "inputs": [],
34
+ "outputs": []
35
+ },
36
+ "text_decoder": {
37
+ "inputs": [],
38
+ "outputs": []
39
+ },
40
+ "token_embedding": {
41
+ "inputs": [],
42
+ "outputs": []
43
+ },
44
+ "vision_encoder": {
45
+ "inputs": [],
46
+ "outputs": []
47
+ }
48
+ }
49
+ }
50
+ ],
51
+ "size": "e2b",
52
+ "_methods_needing_shapes": [
53
+ "audio_encoder",
54
+ "text_decoder",
55
+ "token_embedding",
56
+ "vision_encoder"
57
+ ]
58
+ }
e2b/vulkan/config.json ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
3
+ "model": "gemma_4",
4
+ "family": "gemma",
5
+ "capabilities": [
6
+ "text-generation",
7
+ "image-text-to-text",
8
+ "audio-text-to-text"
9
+ ],
10
+ "backend": "vulkan",
11
+ "license": "gemma",
12
+ "variants": [
13
+ {
14
+ "file": "gemma_4_e2b_vulkan_8da4w.pte",
15
+ "precision": "8da4w",
16
+ "quantized": true,
17
+ "default": true,
18
+ "methods": {
19
+ "enable_dynamic_shape": true,
20
+ "get_bos_id": 2,
21
+ "get_eos_ids": [
22
+ 1,
23
+ 106,
24
+ 50
25
+ ],
26
+ "get_max_context_len": 2048,
27
+ "get_max_seq_len": 128,
28
+ "get_vocab_size": 262144,
29
+ "use_kv_cache": true,
30
+ "audio_block_k_max": 62,
31
+ "audio_block_k_min": 1,
32
+ "audio_block_samples": 7680,
33
+ "audio_encoder": {
34
+ "inputs": [],
35
+ "outputs": []
36
+ },
37
+ "audio_placeholder_id": 258881,
38
+ "audio_sample_count_max": 476160,
39
+ "audio_sample_count_min": 7680,
40
+ "audio_sample_k": 62,
41
+ "audio_soft_tokens_max": 744,
42
+ "has_ple": true,
43
+ "image_placeholder_id": 262145,
44
+ "text_decoder": {
45
+ "inputs": [],
46
+ "outputs": []
47
+ },
48
+ "token_embedding": {
49
+ "inputs": [],
50
+ "outputs": []
51
+ },
52
+ "vision_encoder": {
53
+ "inputs": [],
54
+ "outputs": []
55
+ }
56
+ }
57
+ }
58
+ ],
59
+ "size": "e2b",
60
+ "_methods_needing_shapes": [
61
+ "audio_encoder",
62
+ "text_decoder",
63
+ "token_embedding",
64
+ "vision_encoder"
65
+ ]
66
+ }
e2b/xnnpack/config.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
3
+ "model": "gemma_4",
4
+ "family": "gemma",
5
+ "capabilities": [
6
+ "text-generation",
7
+ "image-text-to-text",
8
+ "audio-text-to-text"
9
+ ],
10
+ "backend": "xnnpack",
11
+ "license": "gemma",
12
+ "variants": [
13
+ {
14
+ "file": "gemma_4_e2b_xnnpack_8da4w.pte",
15
+ "precision": "8da4w",
16
+ "quantized": true,
17
+ "default": true,
18
+ "methods": {
19
+ "enable_dynamic_shape": true,
20
+ "get_bos_id": 2,
21
+ "get_eos_ids": [
22
+ 1,
23
+ 106,
24
+ 50
25
+ ],
26
+ "get_max_context_len": 4048,
27
+ "get_max_seq_len": 4048,
28
+ "get_vocab_size": 262144,
29
+ "use_kv_cache": true,
30
+ "audio_block_frames": 48,
31
+ "audio_blocks_max": 63,
32
+ "audio_blocks_min": 1,
33
+ "audio_encoder": {
34
+ "inputs": [],
35
+ "outputs": []
36
+ },
37
+ "audio_placeholder_id": 258881,
38
+ "audio_sample_rate": 16000,
39
+ "audio_samples_per_block": 7680,
40
+ "has_ple": true,
41
+ "image_placeholder_id": 262145,
42
+ "text_decoder": {
43
+ "inputs": [],
44
+ "outputs": []
45
+ },
46
+ "token_embedding": {
47
+ "inputs": [],
48
+ "outputs": []
49
+ },
50
+ "vision_encoder": {
51
+ "inputs": [],
52
+ "outputs": []
53
+ }
54
+ }
55
+ }
56
+ ],
57
+ "size": "e2b",
58
+ "_methods_needing_shapes": [
59
+ "audio_encoder",
60
+ "text_decoder",
61
+ "token_embedding",
62
+ "vision_encoder"
63
+ ]
64
+ }